Quads, points and power

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
I am trying to quantify average ppd per KW on quads.
The couple of "power usage sites" Ive seen have few if any quads.
I've done a search here and found there are scattered posts in separate threads.
For example:
09-16-2007, 12:36 AM
[BRO]Alaskan Gawd


My GO quad is running 3.06ghz on stock voltage and draws a whopping 162W while folding. (linux SMP)

So I would appreciate if anyone could contribute to this thread with:
Quad version
M/B
RAM
HD ?
OS
Avg PPD
Power usage based on readings from a Kill-A-Watt or similar.

Extenuating factors - such as:
Running on a pizza box with a USB stick and only a CPU fan and 1 stick of memory, no video card.
or
Well it's actually a Xeon server with a 14TB raid and dual high end graphics cards.

Thanks in advance.
Fold ON!
 

aldamon

Supreme [H]ardness
Joined
May 24, 2000
Messages
6,600
I don't consider 162W to be "whopping." That's only 81W per set of cores in the quad. That's comparable to any E6600 I've seen. I don't understand the reaction to the power usage.

Since the quad is the most efficient SMP folder and SMP units give the most points I don't see how it can be beaten in points/watt.

If you really want to make the quad shine in points/watt do this.

no video card...... and dual high end graphics cards.

IMO, video cards are the real power hogs in a system because they're just sitting there drawing power while doing nothing. They're also going to cause the most variability in points/watt results. If you're going to fold with two GPUs installed, then make sure you underclock/undervolt them as much as possible in 2D mode or you're going to be horrified by the power usage. An 8800GTX uses 70W when idle and a 2900xt is off the charts. I use a fanless X300 and a fanless 7300LE in my home folding rigs.


 

ICE_9

2[H]4U
Joined
Feb 28, 2005
Messages
3,459
Since the quad is the most efficient SMP folder and SMP units give the most points I don't see how it can be beaten in points/watt.


I am getting 23.3 P/W on a t7200 system. How does the quad stand up to it?
 

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
I don't consider 162W to be "whopping." That's only 81W per set of cores in the quad. That's comparable to any E6600 I've seen. I don't understand the reaction to the power usage.

Since the quad is the most efficient SMP folder and SMP units give the most points I don't see how it can be beaten in points/watt.

If you really want to make the quad shine in points/watt do this.



IMO, video cards are the real power hogs in a system because they're just sitting there drawing power while doing nothing. They're also going to cause the most variability in points/watt results. If you're going to fold with two GPUs installed, then make sure you underclock/undervolt them as much as possible in 2D mode or you're going to be horrified by the power usage. An 8800GTX uses 70W when idle and a 2900xt is off the charts. I use a fanless X300 and a fanless 7300LE in my home folding rigs.



1. That was a direct quote, perhaps the poster was being sarcastic.
2. I'm not attempting to show "how it can be beaten in points/watt", I.m trying to get a wide sample of points per watt.
3. Nice thread, good figures on PPD, but no power consumption
Thanks for the contribution.

Now does anybody have numbers with power consumption, etc.?

 

aldamon

Supreme [H]ardness
Joined
May 24, 2000
Messages
6,600
Now does anybody have numbers with power consumption, etc.?

http://www.anandtech.com/printarticle.aspx?i=3066

Page 2
Power Consumption: A Cooler SLACR?

I am getting 23.3 P/W on a t7200 system. How does the quad stand up to it?

On what work units? Is that just the CPU or the whole system?

I don't have a Kill-a-watt, but based on this calculator:

http://web.aanet.com.au/SnooP/psucalc2.php

G0 is 5% more energy efficient than B3 in the calculator, so:

2653/3.5 GHz/1.4V/4500 PPD/216.6W = 20.77 PPW

My file server B3 gets

2653/3.0 GHz/1.34V/3641 PPD/187W= 19.475 PPW

Neither machine is barebones so I could probably push that up with some tweaking.


 

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
Thanks, those are some meaningful numbers at anandtech.
They show TOTAL system power consumption as 260W, some of course for the video card.
That's why I'm asking for real world numbers

/edit
One of the reasons I'm asking is that every 100 watts costs me about $60/year for the base watts. In the season I have to cool the house it probably costs twice as much, when I have to heat the house it offsets the gas heat.
Unfortunately I'm still cooling the house (it's 91 deg F outside here in DC). It's getting so cooling season is just December thru April :(

So if I replace an AMD 2700 sucking about 110W with a C2Q sucking 168 watts it's a no brainer (what, about 8 times the power and 20 times the points).

 
Joined
May 14, 2005
Messages
954
http://www.anandtech.com/printarticle.aspx?i=3066

Page 2
Power Consumption: A Cooler SLACR?



On what work units? Is that just the CPU or the whole system?

I don't have a Kill-a-watt, but based on this calculator:

http://web.aanet.com.au/SnooP/psucalc2.php

G0 is 5% more energy efficient than B3 in the calculator, so:

2653/3.5 GHz/1.4V/4500 PPD/216.6W = 20.77 PPW

My file server B3 gets

2653/3.0 GHz/1.34V/3641 PPD/187W= 19.475 PPW

Neither machine is barebones so I could probably push that up with some tweaking.


This is inline with my results
The "whopping" was sarcastic. I consider 162W/hr and 3500PPD (for 1760pt WU's) to be excellent. Mine has a GA-265P-DS3 MB, 2gig (4x 512) DDR2-800, 1 Hard drive, 1 DVD Burner, an X-300 video card (very low power) and the Q6600 (Go)is overclocked to 3.0ghz on stock voltage for 21.6 PPD/watthr. I replaced 3 X2 4200's (110w and 850PPD each) with 2 quads (second one comes online tonight) (162W and 3500PPD ea) approx the same power but triple the points.

EDIT: I forgot to add the EA-430 powersupply
 

ICE_9

2[H]4U
Joined
Feb 28, 2005
Messages
3,459
On what work units? Is that just the CPU or the whole system?


T7200
Asus N4L-VM DH mobo
onboard video
hard drive
onboard NIC
Stock cooler with mobo
Antec EA-500 80+certified power supply
60 watts total measured with a Kill-a-watt
Bench marks from FAHmon

WU 2605
-- SMP T7200 --

Min. Time / Frame : 18mn 04s - 1402.80 ppd
Avg. Time / Frame : 32mn 49s - 772.29 ppd
Cur. Time / Frame : 18mn 05s - 1401.51 ppd
R3F. Time / Frame : 18mn 05s - 1401.51 ppd
Eff. Time / Frame : 18mn 11s - 1393.80 ppd

WU 2653
-- SMP T7200 --

Min. Time / Frame : 17mn 45s - 1427.83 ppd
Avg. Time / Frame : 20mn 36s - 1230.29 ppd
Cur. Time / Frame : 18mn 07s - 1398.93 ppd
R3F. Time / Frame : 18mn 04s - 1402.80 ppd
Eff. Time / Frame : 18mn 06s - 1400.22 ppd
 

Bill Clo

Gawd
Joined
Jan 24, 2002
Messages
959
Quad version Q6600 G0 SLACR @2.8ghz
M/B Asus P5B
RAM 2GB
HD Cheap 80gb
Cheapo PCI vid card
OS Win XP Pro
Antec Earthwatts 380w PSU (83% efficient).
Avg PPD 2550.
Power usage based on readings from a Kill-A-Watt or similar: 147w running SMP client.

Stock Q6600 was 137w I think...
 

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
Quad version Q6600 G0 SLACR @2.8ghz
M/B Asus P5B
RAM 2GB
HD Cheap 80gb
Cheapo PCI vid card
OS Win XP Pro
Antec Earthwatts 380w PSU (83% efficient).
Avg PPD 4550.
Power usage based on readings from a Kill-A-Watt or similar: 147w running SMP client.

Stock Q6600 was 137w I think...
Darn - get a quad, save power!
Now what if you get rid of that hard drive and use a USB drive? And what do you need video for?

 

Bill Clo

Gawd
Joined
Jan 24, 2002
Messages
959
Darn - get a quad, save power!
Now what if you get rid of that hard drive and use a USB drive? And what do you need video for?


You know, I looked into getting a laptop drive that consumed about 2w of power vs 7 for a desktop one, but the payback time was way longer than I expected.

I don't know how to run Windows on a USB drive. Video is a 10w approx cheapo board. Don't know how I can go any lower there.
 

ICE_9

2[H]4U
Joined
Feb 28, 2005
Messages
3,459
I need to look into getting more efficient PSUs. It seems like a worthy investment.



Yep. Those power supplies go on sale at Fry's for $50. I also like the Enhance power supplies also. All my systems are on 80+ power supplies now.
 

Xilikon

[H]ard|DCer of the Year 2008
Joined
Oct 12, 2004
Messages
14,850
[BRO]Alaskan;1031515599 said:
I was going to ask you to post your setup for 4550ppd.

Me too :p I was scratching my head wondering how a setup similar to mine can do 1200 more ppd :eek:

 

MixManSC

║▌║█║▌│║▌║▌█
Staff member
Joined
Aug 12, 2004
Messages
7,225
The box in my sig - I've checked it but I forgot the exact numbers. At true idle I it draws around 300 watts - with dual SMP clients running its around 410 watts. There was another thread on this a while back. I'll find it and give a link in a minute.

http://www.hardforum.com/showthread.php?t=1187754

 

ICE_9

2[H]4U
Joined
Feb 28, 2005
Messages
3,459
My bad, my PPD is 2550 average, not 4550.

I was about to drop some cash on that last night. That would be a folder's dream machine. I couldn't figure out in my head how a quad was beating a 8-way box.
 

jws2346

[H]ard|Gawd
Joined
Feb 28, 2007
Messages
1,768
I've only been folding for about a year, so please ignore these question if they're too dumb. The new penryn C2Ds, as I understand it, will have 6 MBs L2 cache and the penryn quads will have 12 MBs L2 cache (all but the Q9300 with only 6 MBs L2 cache, plus all will have 1333 FSBs) I've also heard or read somewhere about the penryn cache being faster or something.
Way back (in computer terms < than a year) when I got a E6300 cpu and a E6400 cpu, the E6300 was a Conroe with 2 MBs L2 cache working and 2 MBs L2 cache disabled and the E6400 was a true Allendale with only 2 MBs L2 cache.
I went with a E6420 cpu (4 MBs L2 cache) and what a difference 2 MBs of L2 cache made (same boards, same RAM, same WU's, etc), do you think the penryns increased L2 cache and with the new 3 series chipsets, will make that big a difference for ppd? (or the 1333 FSB)
I've read where the Kentsfield Q6600 was a dyamite folder (either B3 or G0)
I've also had an Athlon X2 4200 (only OC'ed to 2.5 GHz or something, I don't remember) and a few single cores folding.
I now fold with a E6420 (OC'ed 2.8 GHz, stock vcore), a E6600 (only mildly OC'ed to 3 GHz) and another E6600 no OC' at all, they fold 24/7. I try to keep my CPU temps down due to folding 24/7.(low vcore, mild OC's) I would appreciate any answer from someone that's into folding. Thanks :D



PS: I also apologize if this is off topic :(
 

schenks

[H]ard|Gawd
Joined
Jun 12, 2006
Messages
1,642
simply put, more cache will pretty much always help. if the processor can store more data locally, rather than having to access ram for the data, then it will be ridiculously faster. even a few megs here and there will help dramatically (as you saw with the 6420), generally as long as the program was already a larger memory amount than the cache size of the old processor.

 

jws2346

[H]ard|Gawd
Joined
Feb 28, 2007
Messages
1,768
Wow, thanks for the fast reply schenksmill. Yeah, it's too bad you can't just add more L2 like you can regular memory (L3?) :)

 

aldamon

Supreme [H]ardness
Joined
May 24, 2000
Messages
6,600
Well, looks like I was off with my calculations. I just got an APC BX1300LCD 780W UPS for my two quad rigs and here's what I'm getting with 100% load on both of them:

dualquadpower.jpg


That's about 15 PPW, not 20 like I reported before.

 

Killer[MoB]

[H]ard|DCer of the Month - March 2008
Joined
Jan 15, 2001
Messages
3,943
Well, looks like I was off with my calculations. I just got an APC BX1300LCD 780W UPS for my two quad rigs and here's what I'm getting with 100% load on both of them:

That's about 15 PPW, not 20 like I reported before.

I picked up one of those from best buy yesterday. 10% coupon and $5 in rewards certs and it was $110 out the door. Not a bad deal at all.

One thing i noticed was that the watts reported by the UPS are lower than reported by my kill a watt. The UPS reported 216 and the kill a watt reported 245.

My question to you is how are you getting 5:38 a frame on a P2653? I can dip to just under 8min a frame if I cut off yahoo messanger and mirc.



 

aldamon

Supreme [H]ardness
Joined
May 24, 2000
Messages
6,600
Killer[MoB];1031629627 said:
One thing i noticed was that the watts reported by the UPS are lower than reported by my kill a watt. The UPS reported 216 and the kill a watt reported 245.

:confused:

Killer[MoB];1031629627 said:
My question to you is how are you getting 5:38 a frame on a P2653? I can dip to just under 8min a frame if I cut off yahoo messanger and mirc.

Not sure what you're talking about.


 

Killer[MoB]

[H]ard|DCer of the Month - March 2008
Joined
Jan 15, 2001
Messages
3,943
Quad version: Q6600 G0 @ 3.52 (1.320v under load after droop)
M/B: Abit IP35 Pro
RAM: 2 gig Corsair XMS
HD: 320 gig Segate SATA 300
Optical: DVD burner
Floppy: Yes
OS: Win XP Pro
Video: Diamond X1950 Pro 512 MB
PSU: Corsair 620w
PPD: (P2653) 3148
Kill-A-Watt: 245w
12.84 PPW

Extenuating factors -
5 - 120mm fans
1 - 200mm fan
USB devices the system is powering - Keyboard, mouse, card reader, fingerprint reader, bluetooth adapter, sidewinder gamepad.

Obviously not a dedicated folding box.
 

Killer[MoB]

[H]ard|DCer of the Month - March 2008
Joined
Jan 15, 2001
Messages
3,943
Not sure what you're talking about.

You said in a earlier post "2653/3.5 GHz/1.4V/4500 PPD".

To get 4500 PPD, you would have to have a frame time of 5:38. I was just wondering what I was missing. A long time ago, computers had TURBO buttons. :eek: DId you figure out how to install one on a new system? I would love to shave 2:25 off my times.
 

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
17,062
Killer[MoB];1031629627 said:
One thing i noticed was that the watts reported by the UPS are lower than reported by my kill a watt. The UPS reported 216 and the kill a watt reported 245.



I was stupidly going to say that the difference would be from the efficiency conversion from the PSU. Luckily, I unfogged what few brain cells I have left before I actually said that since both would be registering the wattage before it reaches the system.

The only thing I can think of is that the wattage meters are just reading differently. I would probably go with the readings reported by the kill a watt rather than the UPS. The kill a watt has only one function and it should do that well. Whereas, the UPS is doing more than one thing and it may not report as accurately since it's a multifunction device.

The only other thing I can think of is if you have something else plugged in that's on the kill a watt and isn't plugged into the UPS.

 
Top